This paper developed a real options model of investment in education and applied it to an ex ante evaluation of the potential impacts of information interventions aimed at improving the accuracy of expectation.
Using data from the Cape Area Panel Study in South Africa, it showed the potential policy relevance of information interventions for at least a subset of indi- viduals, by identifying substantial differences between the expected and observed total value of grade 12 among African youth (though not for Coloured and White youth). This difference is due mainly from differences in the expected and ob- served pathway value of grade 12 (that is, the value of the option to pursue higher education) for African youth.
The impacts of information provision regarding the returns to education were then considered using the Real Options Pathways for Education (ROPE) model, considering the issues of education investment options and heterogeneity in the re- turns to education that received less attention in the existing interventions. Firstly, it was found that information that does not account for possible heterogeneity in the returns to education or higher education options (the simple intervention), would do very little to reduce African youths’ underestimation of the net value of grade 12. Providing information on returns to grade 12 (but not on higher edu-
cation) according to race (the targeted intervention) would not seem to be more helpful than the simple intervention, at least for Africans (and Whites) who, as noted, undervalue the options open with the completion of higher education. The standard intervention, which provides information on all wage outcomes (and, as such, addresses this limitation) but does not consider heterogeneity of returns, would potentially be greatly harmful leading to significant underestimation by African and White youth (35% and 18%, respectively). Lastly, results for the full intervention demonstrate that even when both heterogeneity and the sequential nature of investment are considered, the intervention may lead to a substantial increase in underestimation, the extent of which depends on how new information is used by youth to update their prior expectations.
These results should be seen as a cautionary note regarding the potential and shortcomings of information interventions. To our knowledge, this paper is the first to demonstrate this possibility using real data on observed outcomes and expected outcomes. Similar analyses to this one could be conducted prior to any information intervention where option values and/or heterogeneity in returns are important to evaluate, ex ante, the potential limitations of different types of information provision. Lastly, it must be emphasised that the main contribution of this paper was not to investigate the accuracy of earnings expectations in South Africa and suggest policies for that particular context, but rather to demonstrate a specific analytical approach to planning and evaluating information interventions ex ante. In particular, this approach allows interventions to be checked for their potential to reduce the accuracy of expectations, and in doing so, potentially reduce the efficiency of economic choices.
Social Networks and Wages in South
Africa
3.1
Introduction
An extensive literature in economics and other social sciences, recently reviewed in Marsden and Gorman (2001), Ioannides and Loury (2004) and Topa (2011), analyses the importance of social networks for the functioning of labour markets.1
One of the stylised facts that emerges from this work is that their use is pervasive and generally productive. As a result, a large fraction of jobs are found through social networks. 2
Given this empirical regularity, an important economic question that follows is whether the use of social networks affects labour market outcomes, in particular wages and tenure, and if they do, for whom. As Mouw (2003, p. 868) notes, “This
1The use of social networks refers to job seekers communicating with friends,
relatives and acquaintances about employment opportunities and firms using in- formal referrals and recommendations to find workers.
2This is true in both developed and developing countries. See, for example,
Bewley (1999) who finds that, in the USA, 30-60% of the jobs are found through social contacts. See also Granovetter (1974) and Holzer (1987) for the USA and Bentolila and Suarez (2010) for a recent analysis of the role of social networks in several countries of the European Union. In developing countries see Magruder (2010) and Burns, Godlonton, and Keswell (2010) for evidence in South Africa, Calvo-Armegnol and Zenou (2005) for an analysis of the role of social networks in the functioning of the labour market in Egypt, and Lawler et al. (1995) for a comparison of human resource management practices of firms in India and in Thailand and the extent of use of internal referrals to fill different jobs.
is no trivial issue. If using contacts seems to have little overall impact on labour market outcomes, then perhaps economic models of the labour market can safely ignore ‘embeddedness’ . . . without sacrificing explanatory power.”
The growing number of studies, both theoretical and empirical, that address this question can be organized around two competing hypotheses. The first is the “good matches” hypothesis, which dates back to a seminal study by Rees (1966). More formally, Montgomery (1991), Saloner (1985) and Simon and Warner (1992), interpret social networks as screening or monitoring mechanisms that increase the quality of the match between worker and firm by reducing informational asymme- tries, leading to higher productivity which would be reflected by a higher wage. Alternatively, the “limited choices” hypothesis, initially suggested by Loury (2006), interprets social networks as mechanisms that allocate workers to jobs at a rela- tively low cost for both firms and prospective workers. Because search through networks is limited by network characteristics, including the job information that its members have access to, this type of search may lead to sub-optimal matches and associated wage penalties.
We briefly review both of these different interpretations in section 3.2. We emphasise that they share one important similarity: that the importance of the referee, both in terms of motivation and identity, is usually unexamined.3 Rather, a simple assumption is usually made that “present workers tend to refer people like themselves . . . ” (Rees, 1966, p. 562). Although the importance of homophily, that “birds of a feather flock together”, has long been recognized in the social network literature (see McPherson, Smith-Lovin, and Cook, 2001, for a review), these two competing hypotheses emphasize different aspects of similarity between members
3A referee is someone that refers an individual to a job and/or refers a job to
of a network: productive ability in the first case, identity in the second.
In section 3.2 we also discuss some of the recent work that addresses the possible conflict between these two dimensions of the referee’s decision. We also raise the possibility of referee opportunism, that is, that referees may indicate someone who is not of high ability but who is socially close. An intuitive consequence of this possibility that we explore in this paper, is that firms may react by applying a wage penalty that is a positive function of how close the referee is to the worker. This prediction is close in spirit, to the important study by Granovetter (1974) which emphasises the positive impact of increased social distance (weak ties) between referee and job seeker, on the probability of finding a job.
Focusing on the referee and their incentives is important because previous re- search has found it to impact on the type of person connected to the job and the quality of their work (e.g. Magruder and Beaman (2012)). We do not directly observe the incentives of the referee but assume that they are linked strongly to the identity of the referee (which is observed in our data). Particularly, if the difference in social proximity between friends and relatives is systematic and in the direction we assume (that relatives are closer than friends), we can infer that relatives have a greater incentive to risk referring unsuitable workers.
We empirically estimate the effects on wages of finding a job through the use of different social contacts using data from the Cape Area Panel Study (CAPS), which we present, succinctly, in section 3.3. The CAPS data is particularly valu- able for this purpose, for two reasons. The first is the wealth of information in CAPS on respondents’ employment outcomes, job search methods, and personal characteristics such as education and working experience, for a large sample of youth across several years (2002-2006). The second is the importance of the func-
tioning of the labour market in addressing issues of poverty and inequality in the South African context, a point that we also discuss, briefly, in section 3.3.
In section 3.4 we present our results. We find that, while controlling for indi- vidual and job characteristics, including unobserved heterogeneity, finding a job through social networks carries a wage penalty for all race groups. This suggests that the “good matches” hypothesis finds no support in the context of South Africa. This negative effect is compounded by social proximity, especially in the case of African youth, with higher penalties associated with the use of contacts who are socially closer. 4
We conclude in section 3.5. An understanding of the significance of social networks to individual job search outcomes may have important policy implications in the high unemployment context of South Africa. If youth are using social networks to acquire jobs even though they are associated with large wage penalties, this may be a sign that labour market institutions do not facilitate formal job search for these individuals to find more suitable employment.